List your company

Object Hallucination Probing on GQA POPE Random

89.93Accuracy (GQA POPE)Scalpel
57.6966.0674.4382.8Dec 29, 2025Jan 14, 2026Jan 31, 2026Feb 17, 2026Mar 6, 2026Mar 23, 2026Apr 9, 2026

Evaluation results

MethodLinks
ScalpelBase Model=LLaVA-1.5-7B2026.0289.93—————————89.87——
MESAModel=LLaVA-v1.52026.0489.5—————————89.459089
ICTBase Model=LLaVA-1.5-7B2026.0289.3—————————89.49——
ICTModel=LLaVA-v1.52026.0489.26—————————88.7892.9385
CoFi-DecBackbone=LLaVA-1.52025.1289.03—————————89.0681.1—
VAFModel=LLaVA-v1.52026.0489—————————88.6391.6685.8
VTIModel=LLaVA-v1.52026.0488.37—————————88.0290.7385.47
OPERABase Model=LLaVA-1.5-7B2026.0288.13—————————88.91——
MESAModel=Qwen-VL2026.0487.97—————————86.6797.1178.27
NulluModel=LLaVA-v1.52026.0487.7—————————8885.990.2
CoFi-DecBackbone=Qwen-VL2025.1287.14—————————86.3294.65—
DeGFBackbone=LLaVA-1.52025.1287.09—————————87.9680.46—
CoFi-DecBackbone=InstructBLIP2025.1286.78—————————86.3987.06—
VCDBase Model=LLaVA-1.5-7B2026.0286.65—————————86.99——
VCDModel=LLaVA-v1.52026.0486.65—————————86.9984.8589.24
ICDModel=Qwen-VL2026.0486.2—————————84.6295.5575.93
VAFModel=Qwen-VL2026.0486.11—————————86.1196.8777.53
RITUALBackbone=LLaVA-1.52025.1286.1—————————87.3180.3—
RITUALBackbone=Qwen-VL2025.1286.1—————————84.8193.78—
DeGFBackbone=Qwen-VL2025.1285.95—————————85.0894.22—
ICDModel=LLaVA-v1.52026.0485.9—————————85.2989.181.8
M3IDBackbone=Qwen-VL2025.1285.69—————————84.6793.11—
VCDModel=Qwen-VL2026.0485.59—————————85.3386.8883.84
DeGFBackbone=InstructBLIP2025.1285.4—————————85.1285.64—
ICTModel=Qwen-VL2026.0485.33—————————86.1295.9578.13
VCDBackbone=Qwen-VL2025.1285.21—————————84.2192.05—
VanillaModel=LLaVA-v1.52026.0484.96—————————86.2579.4494.33
RegularBackbone=Qwen-VL2025.1284.9—————————83.9689.51—
RITUALBackbone=InstructBLIP2025.1284.87—————————85.3982.52—
VTIModel=Qwen-VL2026.0484.33—————————81.9496.7371.07
M3IDBackbone=LLaVA-1.52025.1284.2—————————85.7778—
VanillaBase Model=LLaVA-1.5-7B2026.0283.73—————————82.95——
VCDBackbone=LLaVA-1.52025.1283.23—————————85.0576.73—
M3IDBackbone=InstructBLIP2025.1283.07—————————83.8780.06—
VCDBackbone=InstructBLIP2025.1282.83—————————83.5680.16—
RegularBackbone=LLaVA-1.52025.1282.23—————————84.0376.32—
RegularBackbone=InstructBLIP2025.1279.67—————————80.9976.05—
VanillaModel=Qwen-VL2026.0474.76—————————66.698.4350.33
NulluModel=Qwen-VL2026.0474.73—————————65.5598.4350.26
VCDModel=Qwen-VL, Decoding=VCD2026.0159.7—————————56.4361.4152.2
SDCDModel=Qwen-VL, Decoding=SDCD2026.0158.97—————————55.3560.750.87
Regular DecodingModel=Qwen-VL, Decoding=Regular2026.0158.93—————————54.4461.1349.07
BaselineBackbone=LLaVA-1.52025.06—89.4777.0363.8370.2374.8768.07——70.81———
BaselineBackbone=Qwen-VL-Chat2025.06—87.2382.5373.0374——65.174.373.95———
CLAIMBackbone=LLaVA-1.52025.06—84.2785.1383.5782.5382.67———83.63———
CLAIMBackbone=Qwen-VL-Chat2025.06—85.1779.9381.962.03——80.4—77.89———
VCDBackbone=LLaVA-1.52025.06——7868.3371.4773.6370.2——72.33———
VCDBackbone=Qwen-VL-Chat2025.06——83.6376.173.43——55.580.273.77———